【作者】 陈飞； 王秀芳； 农宇；

【Author】 CHEN Fei~1,WANG Xiu-fang~2,NONG Yu~1 (1.School of Resource and Environment Sci.,Wuhan Univ.,Wuhan 430079,China; 2 School of Remote Sensing and Info.Eng.,Wuhan Univ.,Wuhan 430079,China)

【机构】 武汉大学资源与环境科学学院； 武汉大学遥感信息工程学院；

【摘要】 针对彩色扫描地图的点状符号,提出了一种基于距离变换进行模式识别的新方法。首先进行符号的提取,对地图进行分色提取分版图,引入线划弯曲度的概念,根据弯曲密度和线划密度过滤掉非点状符号的线划,并利用距离变换的方法提取点状符号。然后进行符号的识别,通过搜索符号与模板之间的最佳匹配点来确定待识别的符号区域,提取符号的骨架线并由骨架线上的内距值来判断符号的拓扑连接结构,为了更好的描述图像间的相似性提出了加权距离函数。试验证明,该方法能适用于背景复杂情况下的旋转或变形的符号,识别准确率达到满意的程度。更多还原

【Abstract】 It presents a new method of recognizing point-shaped symbols of color scanned maps based on Distance Transformation.First,it extracts symbols.It picks up element map by separating colors,introduces the conception of line camber,and filters the lines of non-point symbols according to curve density and line density,and extracts point symbols from the map based on distance transformation.Second,it identifies the symbols.It determines symbol region by searching for optimal matching points between symbols and templates,extracts skeleton lines of symbols and judges symbol topological structure by inner distance of skeleton lines,and weighted distance function is presented to describe the similarity between images better.The experiment shows that this algorithm can be applied to identify rotated or distortional symbols in the condition of complex background,and the identification accuracy is satisfactory.更多还原